Abstract
Function–structure connectivity (FSC) coupling provides an interpretable perspective to characterize how the relationship between brain functional connectivity (FC) and structural connectivity (SC) is altered in neurodegenerative conditions. However, most existing FSC coupling approaches treat functional MRI (fMRI) signals as discrete observations, failing to explicitly capture continuous neural dynamics underlying brain activities. Although graph neural ordinary differential equations (Graph ODEs) enable continuous-time modeling of latent neural state evolution, they generally ignore the structural pathways that constrain functional neural signal propagation. To address these limitations, we propose a novel Anatomy-constrained Neural Dynamics (AND) learning framework, integrating FSC coupling with continuous-time modeling for cognitive decline analysis. Specifically, regional fMRI time series are encoded into latent neural states and evolved via Graph ODEs. In our framework, functional interactions are softly modulated by physical white matter pathways, yielding structure-constrained message passing dynamics. The resulting latent trajectories are decoded to reconstruct fMRI signals, from which more reliable FC is derived. Based on the reconstructed FC and physical SC, we derive node-wise FSC coupling profiles for downstream cognitive decline detection. Experimental results on the ADNI dataset with paired resting-state fMRI and diffusion tensor imaging (DTI) data demonstrate that our method outperforms state-of-the-art approaches. With the learned coupling representations, the proposed AND can reveal localized FC-SC interaction patterns associated with neurocognitive decline, providing potential imaging biomarkers.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/MLMI_014.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
Open Review Page: https://openreview.net/forum?id=CnoUWbXCLm
BibTex
@InProceedings{WanQia_AnatomyConstrained_MICCAISAT2026,
author = { Wang, Qianqian AND Wu, Mengqi AND Sun, Yongheng AND Liu, Mingxia},
title = { { Anatomy-Constrained Neural Dynamics Learning with Function-Structure Coupling for Cognitive Decline Analysis } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17272},
month = {pending},
page = {pending}
}